How does it integrate with Claude Code?

SLB includes a dedicated 'PreToolUse' hook that intercepts Claude's command execution attempts and routes them through the approval workflow based on the risk tier.

Can I undo a command executed via SLB?

Yes, SLB supports state capture and rollback for many commands, allowing you to restore the filesystem, Git state, or Kubernetes manifests if something goes wrong.

Does it require a central server?

No, SLB uses client-side execution. Commands run in your local shell environment, ensuring they inherit your specific AWS profiles, Kubeconfigs, and SSH keys.

SLB (Simultaneous Launch Button) is a security tool that requires a second approval before an AI agent can execute high-risk CLI commands, similar to a nuclear launch key system.

How does it classify command risk?

It uses a shell-aware pattern-matching engine to sort commands into Critical, Dangerous, Caution, or Safe tiers based on their potential impact on the system.

SLB Secure Command Guard

Name: SLB Secure Command Guard
Author: Dicklesworthstone

byDicklesworthstone

•

安全与测试

Implements a two-person approval rule for destructive CLI commands executed by AI agents to prevent data loss and infrastructure damage.

SLB (Simultaneous Launch Button) is a Go-based security utility designed to provide a critical governance layer for multi-agent coding workflows. It automatically classifies CLI commands into four risk tiers—Safe, Caution, Dangerous, and Critical—enforcing mandatory peer review for high-stakes operations like forced git pushes, database deletions, or infrastructure teardowns. By requiring a second "notary" approval before execution, SLB mitigates the risks of agent hallucinations and tunnel vision while ensuring that commands inherit local shell environments and pass a rigorous five-gate verification process including SHA-256 integrity checks.

主要功能

01Multi-tier risk classification engine for shell commands

0259 GitHub stars

03Secure 5-gate verification process with command hash binding

04Mandatory two-person rule enforcement for destructive agent actions

05Automated dry-run pre-flights for Terraform, K8s, and Git

06State capture and rollback capabilities for emergency recovery

使用场景

01Auditing and logging every potentially destructive action taken by automated coding assistants

02Enforcing peer-review workflows for high-privilege git operations in multi-agent environments

03Preventing AI agents from accidentally deleting production databases or cloud infrastructure

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add dicklesworthstone/agent_flywheel_clawdbot_skills_and_integrations slb

For use in Claude.ai and ChatGPT

Download Skill